Você está na página 1de 10

University Of Nottingham

School Of Economics
MSc Economics of Monetary & Financial Policy

Module: L14009 Economic Data Analysis

Unemployment, labour force participation, per-capita


income, alcohol consumption and suicide: a micro panel
analysis.

By Ayandip Chakrabarti
Student ID - 4310012
Introduction
Over the last few years suicide rate consisting of both male and female for all
age groups has increased dramatically. Suicide is a groundless desire to die. The
term groundless is used because how bad a person’s life is it is a permanent
solution to a temporary problem. A person who commits suicide leaves behind a
confusion of family members and friends who try to make sense of a senseless
act. This paper addresses the effect of alcohol consumption, unemployment,
labour factor productivity, gross domestic product per capita on suicide rate of
both male and female for twenty nine different countries from (1990-1995).The
main purpose is to test whether the factors stated above are significantly related
to suicide rate and to what extent. The rest of the paper follows as, Literature
review, econometric framework and defining the variables, discussing about the
data set, empirical analysis, conclusion.

Literature Review
According to Soss (1974) theory of suicide a person will commit suicide when
the life time discounted utility available falls below a certain threshold. Higher
the income, higher the utility and therefore living is more desirable than to
commit suicide. In contrast Lester (1996) has stated that with increase in
economic development suicide rate tends to increase. Unemployment means
less opportunities available to a person and thereby reduces the person’s income
and increases the chances of committing suicide. However unemployment can
be related to depression, loss of confidence, anxiety that leads to suicide. Recent
micro panel data studies have confirmed that (Chuang and Huang, 1997)
adverse relationship between suicide mortality and unemployment. Taking the
case of (Ruhm) has found a positive effect of U.S. total unemployment rate and
total suicide rate. According to (Durkheim 1951) he founded that suicide rate
are also influenced by social integration and regulation. From this perspective
divorce and fertility rate can be viewed as indicators of social integration.
Divorce leads to reduction in social integration and family ties and countries
having high divorce rate tends to have high suicide rate. A few recent studies on
alcohol consumption and suicide rate (Neumayer 2003, Ramsted 2001) has
stated that alcohol consumption leads to increase in suicide rate. Alcohol and
suicide rate can be related to each other through unobserved factors. Alcoholics
can show other forms of psychological problems like mental disbalence, lack of
friends and little support from society. Therefore high level of alcohol
consumption can be related to the factors stated above and leading to increase in
suicide rate.
Econometric Model
The model used in this paper takes the following form.

LnSRit = β0 +δ1y91t+….δ5y95t+β1(Unit)+β2(ACit)+β3(LFPit)+β4(GDPPit)+µit
The model above estimates whether the explanatory variables listed above are
significantly related to suicide rate. I am going to describe the variables and the
source of data collected from.
Variables and Definitions.
LnSRit is the log of suicide rate , age standardized rate of both male and female
from year 1990-1995, it is used as the dependent variable. Where i represents
country ID and t represents time period.

δ1y91t is the time dummy which has been introduced from the year 1991 to
1995 to check whether suicide rate has increased over time in the pooled ols
estimate.
β1(Unit) is the unemployment rate of both male and female from period 1990-
1995.It is used as the independent variable.
β2(ACit) is the alcohol consumption of both male and female form period 1990-
1995.
β3(LFPit) is the labour force participation rate of both male and female from
period 1990-1995.
β4(GDPPit) is the combined per capita income of both male and female from
1990-1995.
Data Set
Data has been collected from World Bank , OECD , World Health Organization
& World drinks trend .
Emperical Analysis
In this section we start by first summarizing the data set by stating the countries
with highest suicide rate, followed by the (average) alcohol consumption,
unemployment rate and per- capita income of these countries. Then followed by
using graphs, pooled ordinary least squares model with using the cluster robust
standard errors. Then we estimate the model using first differencing estimator,
followed by fixed effect and random effect and finally we perform the hausman
test and check whether the p value is significant or not and null hypothesis is
rejected or not. Now in the following table we describe the summary of 10
countries with highest suicide rate but in the data there are 29 countries we have
taken the highest ones.

Country Suicide Rate Alcohol Unemployment Per-capita


Consumption Rate
Russia 33.50364 8.48 5.808 3054.071
Hungary 31.91631 13.06 8.876 3913.887
Finland 26.07525 8.75 11.7833 23402.89
Austria 19.10496 13.5 4.778 24777.86
Switzerland 17.97715 12.23 2.931667 40725.94
Denmark 17.27603 11.85 8.70333 29462.9
France 17.57737 15.64 10.74167 23446.91
Belgium 16.99914 11.51667 8.001667 23445.01
Czech 16.49654 11.533 3.035 4062.016
Republic
Sweden 13.50686 6.28333 6.4333 28936.23
Pooled Ordinary Least Square Estimate
LnSR Coef. Std Er t P>|t| [95% Conf. Interval
y 91 .0569188 .154345 .37 .713 -.2480858 .3619233
y 92 .0283721 .1517653 .19 .852 -.2715347 .3282788
y 93 .002207 .1509403 .01 .988 -.2960695 .3004835
y 94 .045763 .1530292 .30 .765 -.2566415 .3481674
y 95 .0458963 .1511046 .30 .762 -.252705 .3444975
GDPP 8.12e-06 4.50e-06 1.81 .073 -7.69e-07 .000017
LFP .341833 .0073204 4.67 .000 .0197174 .0486493
AC -.0005242 .0001442 -3.63 .000 -.0008092 -.0002391
UN .024568 .0108293 2.27 .025 .003168 .045968
CONS .0035354 .4604691 .01 .994 -.9064079 .9134787

Sources SS df MS
Model 14.1615555 9 1.57350617
Residual 40.9279921 148 .276540487
Total 55.0895476 157 .350888839

F(9, 148) = 5.69, Prob> F = .0000 R –squared =.2571, Adj R –squared= .21119, Root MSE = .52587

With the performing of the pooled ordinary least square estimate, it can be seen
that suicide rate has increased over the years and it has been positive , however
there was little bit reduction in the suicide rate in year 1993. When analysing
the p values it states that the values are greater than the .05 so it is not
significant and cannot conclude that a significant difference exists. Since a log
level model is used, while estimating the coefficients unemployment shows that
a unit increase in unemployment increases suicide rate by 2.4568%.The p value
associated with unemployment is .025 and it is significant. Alcohol
consumption shows that with a unit increase in alcohol consumption decreases
suicide rate by .05242% and the p value is .000 we can state it is significant.
Labour factor productivity shows that with a unit increase in labour factor
productivity suicide rate increases by 34.1833%, now looking into the p value it
can be stated that it is significant since the p value is less than .05.In the case of
gross domestic per-capita income it can be seen that it is highly insignificant as
the coefficient is way too small and looking at the p value since it is greater than
.05 and thus it is insignificant. The overall f test is significant. The value of r
square is .2571. Now in the next part we will use graphs and then use the same
regression with cluster-robust – standard error.
It can be stated from the diagram that the relationship between unemployment
and log of suicide rate is significant with only few outliers. Where as in the case
for log of suicide rate and labour factor productivity it shows a significant
relation as estimated in the regression. With the case for per-capita income and
log of suicide rate it can be stated that there are large no of outliers and as stated
in the regression results that it is not significantly related to suicide rate.

Robust Standard Errors


Ln SR Coef Robst Std Er t P>|t| [95 % Conf. Interval]
y 91 .0569188 .0818178 .70 .492 -.1106773 .2245148
y 92 .0283721 .0835617 .34 .737 -.1427964 .1995405
y 93 .0022207 .087345 .03 .980 -.1767112 .1811252
y 94 .045763 .0929734 .49 .626 -.1446844 .2362104
y95 .0458963 .1154272 .40 .694 -.1905664 .2823589
GDPP 8.12e-06 .0000113 .72 .479 -.0000151 .0000313
LFP .341833 .018419 1.86 .074 -.0035462 .0719128
AC -.0005242 .0001433 -3.66 .001 -.0008178 - .0002306
UN .024568 .0000485 1.82 .035 -.0167044 .0658405
Cons .0035354 1.142627 .00 .998 -2.33703 2.344101

F(9, 28) =28.17, Prob> F = .0000, R- Squared = .2571,Root mse = .52587

Now after performing the pooled ordinary least square we use the robust
standard errors. Now analysing the results, it can stated that for the time dummy
variables the coefficients are similar to the results obtained in the pooled
ordinary least square , there has been increasing in the suicide rate over the
years , however there was decrease in suicide rate in year 1993 but again it has
increased. But although looking into the p values it can be stated that it is not
statistically significant. However it is observed that the t values have become
bigger in comparison to the pooled estimate. It can be stated that for alcohol
consumption, the p value is significant and thus it can stated that with one unit
increase in alcohol consumption suicide rate decreases by .05242%. For per-
capita income the p value is insignificant at 5% significance level, as before and
the coefficient is very small which states that with increase in one unit in per-
capita increases suicide rate by a very negligible amount. For unemployment it
can be stated that with increase in one unit of unemployment increases suicide
rate by 2.4568% and it is statistically significant. Now the labour factor
productivity it can be stated that the p value is significant at 1% significance
level. With one unit increase in labour factor productivity increases suicide rate
by 34.183%. The r square remains the same and the f value is significant.
First Differencing estimator
D.ln SR Coef Std. Err t p>|t| [95% Conf. Interval]
y91 D1 .007299 .0130562 .56 .577 -.018558 .0331561
y92 D1 -.0166434 .0154743 -1.08 .284 -.0472893 .0140026
y93 D3 -.0191153 .0168258 -1.14 .258 -.0524379 .0142072
y94 D1 -.0112789 .0135835 -.83 .408 -.0381802 .0156225
y95 D1 0 (omitted)
GDPP D1 -3.12e-06 3.99e-o6 -.78 .436 -.000011 4.79e-06
LFP D1 .0100894 .0036815 2.74 .007 .0027984 .0173803
AC D1 .0403915 .0152949 2.64 .009 .0101008 .0706823
UN D1 .0035745 .0048585 .74 .463 -.0060474 .0131965
Cons .003847 .0077361 .50 .620 -.011474 .0191679
F(8, 117) = 2.04 , Prob>F = .0434 R squared = .1224, Adj R squared = .0624, Root Mse = .06722.

Sources SS df MS
Model .073746119 8 .009218265
Residual .528622609 117 .004518142
Total .602368728 125 .00481895

Now using the first differencing estimator we can state that unemployment has
become statistically insignificant and it can stated that unemployment is
correlated with other unobserved fixed effects. The coefficient of
unemployment is now also reduced. It can be stated that the coefficient for the
unemployment is now .0035745 lower compared to the coefficient .024568 in
the ols regression. However the alcohol consumption is now statistically
significant and the labour factor productivity is significant at 5 % significance
level. However the per- capita remains insignificant. The overall f test is
significant.

Fixed Effect Estimator


Ln SR Coef. Std. Err t P>|t| [95% Conf. Interval]
y91 .0092064 .0206402 . 45 .656 -.0316597 .0500726
y92 -.0184963 .0220877 -.84 .404 -.0622283 .0252357
y93 -.0115758 .0224739 -.52 .607 -.0560727 .032921
y94 -.000251 .0241359 -.01 .992 -.0480384 .0475364
y95 .012908 .0285574 .45 .652 -.0436337 .0694497
GDPP -1.52e-06 3.53e-06 -.43 .668 -8.50e -06 5.47e-06
LFP .014599 .0040174 3.63 .000 .006645 .0225531
AC .0309444 .0113773 2.72 .008 .008418 .0534707
UN .0019675 .0042318 .46 .643 -.0064111 .0103461
Cons -.3224402 .7361335 -.44 .662 -1.779933 1.135053
Sigma_u 11.612409
Sigma_e .06846237
rho .99996524
No of groups = 29 R-sq within = .1658 Observation per group: min = 1,
avg = 5.4, max = 6 between=.0809, overall= .68 F (9, 120) = 2.65, Prob F> .0 078

Random Effects Estimates


Ln SR Coef. Std. Err z P>|z| [95% Conf. Interval]
y91 .0053922 .0227696 .24 .813 -.0392354 .0500198
y92 -.0329889 .0238885 -1.38 .167 -.0798096 .0138318
y93 -.0267495 .0243523 -1.10 .272 -.0744792 .0209802
y94 -.0205285 .025719 -.80 .425 -.0709349 .0298778
y95 -.0109883 .029957 -.37 .714 -.0697029 .0477263
GDPP 9.64e-07 3.55e -06 -.27 .786 -6.00e-06 7.92e-06
LFP .0156059 .0042647 3.66 .000 .0072472 .0239645
AC -.0005276 .0002439 -2.16 .031 -.0010057 -.0000495
UN .0033845 .0045127 .75 .453 -.0054603 .0122293
Cons 1.454591 .2785474 5.22 .000 .9086477 2.000533
Sigma _u .42440594
Sigm_e .06846237
rho .97463796
No of obs = 158, No of groups = 29 R- sq within = .1093, between =.1707, overall = .1989
Observation Per group: min = 1, avg = 5.4, max = 6 , Wald chi2(9) = 21.07, Prob> chi2 = .0123

Now analysing the results for random effects and fixed effects. For the fixed
effect it can be stated that the unemployment remains statistically insignificant,
alcohol consumption and labour factor productivity remains statistically
significant because the p value is than .05 but per-capita remains insignificant.
The standard deviation of the fixed effects is 11.612409 and the standard
deviation of the error term is .06846237. Now comparing the p value of random
effect and fixed effect. It can be stated p values for per-capita and
unemployment remains statistically insignificant.

Hausman Test
fe re Difference S.E.
y91 . 0092064 .0053922 .0038143 -
y92 -.0184963 -.0329889 .0144926 -
y93 -.0115759 -.0267495 .151737 -
y94 -.000251 -.0205285 .0202775 -
y95 .012908 -.0109883 .0238963 -
GDPP -1.52e-06 9.64e-07 -2.48e -06 -
LFP .014599 .0156059 -.0010068 -
AC .0309444 -.0005276 .013472 .0113747
UN .0019675 .0033845 -.0014169 -
Prob>chi2 = .8721 chi2(8) =3.83
Now we finally perform the Hausman Test to check which estimator can be
used as the preferred estimator. It can be seen that the difference between the
fixed effect and random effect is very less we cannot reject the null hypothesis
since the p value is greater than .05 and it can be stated that both fixed effect
and random effects are consistent estimators.

Conclusion
It can be stated that from performing the hausman test we cannot reject the null
hypothesis and fixed effect and random effect are consistent estimators. We can
state that alcohol consumption and labour factor productivity are statistically
significant to suicide rate, and thus we can state that with increase in
unemployment rate and increase in alcohol consumption the age standardized
suicide rate will increase. But per-capita is insignificant with suicide rate and as
stated earlier that unemployment rate is correlated with unobserved fixed effect
and hence it is statistically insignificant.

References
• Neumayer, E. Socio- economic factors and suicide rate.
• Chuang, H, L- Economic and social correlates of regional suicide rates.
• Antonio Andres- Income inequality and suicide a panel analysis of 15
European countries.

Você também pode gostar